# GEO agency for ChatGPT and Perplexity

How AI engines choose which tools to name: query fan-out, the pages they read and share of answers, measured. GEO for B2B SaaS. We automate it.

Canonical: https://ignura.com/ai-search
Last updated: 2026-10-27
Author: Ignura

## In short

- GEO means getting your product named when a buyer asks an AI engine which tool to use. It’s a separate fight from Google: of the 80 pages ChatGPT and Perplexity cited for our test questions, only 3 were on Google’s first page.
- The engines don’t answer from memory. For one buyer question ChatGPT ran 4 web searches, read what they found, then named five tools.
- So you get named by being on the pages the engines read, and by having pages of your own that answer the question in their first lines.
- We measure your share of answers on your 20 buyer questions every week and fix what keeps you out. Start with a free Growth Check.

## How an engine picks who to name

ChatGPT search, Perplexity, Claude and Gemini work the same way at the top: the model turns the question into web searches, reads a handful of pages, writes an answer and cites some of what it read. Google calls this query fan-out and says AI Overviews and AI Mode use it. We watched ChatGPT do it: 15 searches for our 6 test questions.

That makes the pages it reads the real ballot. If the comparison pages a search returns all list your rivals, the answer will too. If none of them mention you, polishing your own homepage won’t get you in.

**9 of 15** ChatGPT’s searches went straight to a vendor’s own site, most of them for pricing.

## Two engines, two first picks

Asked the same question on the same day, ChatGPT put Vanta first and Perplexity put Sprinto first. Across six questions the two engines named Vanta and Drata in 10 of 12 answers each, Secureframe in 7 and Sprinto in 6. Each engine has its own search provider, crawler and habits, so we track each one on its own.

Answers also move between runs. SparkToro had volunteers run 12 prompts 2,961 times across ChatGPT, Claude and Google’s AI, and the same list of brands came back less than once in 100. One check tells you little. A share measured weekly over the same questions tells you a lot.

## What the engines can reach

ChatGPT search reads the web through OAI-SearchBot, and OpenAI says sites that block it won’t be shown in ChatGPT search answers. Perplexity uses PerplexityBot to surface and link sites. Anthropic says blocking Claude-SearchBot may reduce your visibility in Claude’s search results. Plenty of companies block these by accident, with a “block AI” rule or a firewall setting copied from a guide.

Most AI crawlers don’t run JavaScript either. Vercel found none of the major ones did in late 2024; only Google’s and Apple’s render it. If your pricing or feature list only appears after a script runs, the engine reads an empty page.

## What gets a page cited

The cited pages had one thing in common: they answered the exact question in their first lines. “Best SOC 2 software for startups”, “Vanta vs Drata”, “Vanta alternatives”. 72 of the 122 citations went to comparisons, “best” lists or alternatives pages.

Your own site counts too. A clear pricing page, a plain line on who the product is for, and one comparison page you’d stand behind are the cheapest wins in GEO.

## Mistakes we see

- **Blocking the search crawlers.** A rule that stops GPTBot only stops training. One that also stops OAI-SearchBot or PerplexityBot takes you out of their answers.
- **Asking once and believing it.** Answers change from run to run. One screenshot of ChatGPT naming you, or not, proves nothing.
- **Writing for a keyword, not a question.** Buyers ask whole questions with their situation in them. A page has to answer that question near the top.
- **Only working on your own site.** Two thirds of what the engines cited were other people’s pages. If you’re not on them, you’re not in the answer.

## What our AI does

- Asks your 20 buyer questions on ChatGPT search and Perplexity every week and records who is named.
- Logs every search the engines run and every page they cite.
- Ranks the pages that name your rivals and not you by how often they decide an answer.
- Checks your robots.txt and firewall rules against every AI crawler.

## What we do

- Pick the 20 questions with you, in your buyers’ words.
- Write and fix the pages that answer them, and put your product in front of the people behind the cited pages.
- Decide what to fix first, and say plainly when something won’t move.
- Answer for the number. We’re AI-native: software does the asking and logging, people do the judging and the writing.

## The first four weeks

1. We agree your 20 buyer questions and measure where you stand on ChatGPT search and Perplexity. We check every AI crawler can reach your pages and read them without JavaScript.
2. We fix your own pages first: pricing, who it’s for, one honest comparison. We list the third-party pages the engines cite and start on the ones that should name you.
3. New pages go up for the questions you’re missing from, and we ask all 20 again to see what moved.
4. We measure the same 20 questions the same way as in week 1. You get every answer, the report and the plan for month 2.

## The deal

Named in 5 of your 20 buyer questions by week 4, or you don’t pay the second half. $8,000 for four weeks, half to start. Then $4,000 a month, with a month’s notice to stop.

## Questions

### Is GEO different from SEO?

Partly. Both need pages that answer real questions and can be crawled. The difference is who reads them. In our test only 3 of the 80 pages the AI engines cited were on Google’s first page for the same question; across 20 questions in our breakdown work it was 2 of 383. Ranking on Google doesn’t get you into AI answers by itself.

### Can you guarantee ChatGPT will recommend us?

No one can promise a single answer, because answers change between runs. We guarantee a share instead: Named in 5 of your 20 buyer questions by week 4, or you don’t pay the second half.

### Which engines do you track?

ChatGPT search and Perplexity every week, because both search the live web and show their sources, so we can see why you were or weren’t named. Google’s AI Overviews are covered on our AI Overviews page. Claude and Gemini we check on the questions that matter most.

### How fast do changes show up?

Changes to your own pages count once the engines read them again; OpenAI says a robots.txt change takes about 24 hours to reach its systems. Getting onto other people’s pages takes longer, because someone else has to update them. That’s why the guarantee sits at week 4.

### Do you add llms.txt?

We can, but it isn't how you get named. It's a 2024 proposal. Google's AI optimization guide says an llms.txt file "will neither harm nor help your site's visibility or rankings in Google Search", and none of OpenAI, Anthropic or Perplexity says its crawlers read it. Our technical SEO page has the detail.

## Sources

- Ignura: [Six SOC 2 buyer questions on ChatGPT search, Perplexity and Google, 27 Oct 2026 (the data)](https://ignura.com/data/search-2026-10-27.json)
- Ignura: [20 B2B buyer questions on ChatGPT search, Perplexity and Google US, 27 Oct 2026 (growth breakdown work)](https://ignura.com/ai-search#measured)
- Google Search Central: [AI features and your website](https://developers.google.com/search/docs/appearance/ai-features)
- OpenAI: [Overview of OpenAI crawlers](https://developers.openai.com/api/docs/bots)
- Perplexity: [Perplexity crawlers](https://docs.perplexity.ai/guides/bots)
- Anthropic: [Does Anthropic crawl data from the web, and how can site owners block the crawler?](https://support.claude.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler)
- Vercel: [The rise of the AI crawler (17 Dec 2024)](https://vercel.com/blog/the-rise-of-the-ai-crawler)
- SparkToro: [AIs are highly inconsistent when recommending brands or products (27 Jan 2026)](https://sparktoro.com/blog/new-research-ais-are-highly-inconsistent-when-recommending-brands-or-products-marketers-should-take-care-when-tracking-ai-visibility/)
- llmstxt.org: [The /llms.txt file (Jeremy Howard, 3 Sep 2024)](https://llmstxt.org/)
- Google Search Central: [AI optimization guide](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide)

Book a 15-minute call about your website: https://ignura.com/book
